Indexicality and De Se Reports

or introduced by an attitude operator, there could in principle be further syntactic constraints on the relation between a De Se pronoun and its antecedent. Percus & Sauerland (2003b) claim that such constraints exist in dream reports, as in (48): (48) a. John dreamed that he was marrying his granddaughter. b. John dreamed that his granddaughter was marrying him. They argue on the basis of intricate scenarios that if he is read De Re in (48)a, his must be read De Re as well; by contrast, even when him is read De Re in (48)b, his can still be read De Se. They conclude that a De Re pronoun pro1 that refers to an attitude holder x can block the De Se reading of the pronoun pro2 in the configuration [x dreams that ... pro1 ... pro2...] if pro1 c-commands pro2 (‘Oneiric Reference Constraint’). This constraint is further discussed by Hardt (2003) and Anand (2006), who suggests that it is an instance of a more general condition (the ‘De Re Blocking Effect’), which has analogous effects on some logophoric pronouns, and ultimately derives from economy constraints on binding discussed in Fox (2000). 3. Monsters and shifted indexicals In this section, we suggest that Kaplan’s analysis was not just wrong about De Se readings, but also about monsters: there are languages in which (some) attitude operators behave like Kaplanian monsters (a conclusion consonant with Israel & Perry 1996). But how can we establish the existence of such beasts? We will discuss examples that have the form of (49), where and are indexicals such as I and here: (49) John says that ... ... ... The argument will have three steps, each of which is compulsory if the claim that a monster has been sighted is to have any substance. (i) First, by inspecting the truth conditions, we argue that this hypothesis is compatible with the semantics of the sentence. Often one only checks that the shifted indexical has roughly the right meaning for instance that in John says that I am a hero the embedded pronoun intuitively refers to John rather than the actual speaker. But within the present framework, a bona fide shifted indexical should obligatorily be read De Se; if possible, then, the De Re / De Se distinction should be considered when shifted indexicals are discussed (this is admittedly difficult when one cannot do detailed fieldwork on the relevant constructions). (ii) Second, we will have to exclude the possibility that the embedded clause is quoted. On almost any theory, it is unsurprising that John says: ‘I am a hero’ attributes to John a (De Se) claim about John himself. This is because in cases of quotation, the verb say can be taken to establish a relation between an individual and a string of words, rather than between an individual and the meaning of the embedded clause (this explains why meaningless strings can be quoted in direct discourse but not used in indirect discourse; thus it makes sense to